Maritime Surveillance Physics

Interactive visualizations of the core physics governing radar detection, underwater acoustic propagation, and passive sonar performance.

Radar Range & Sea Clutter

Interactive radar sweep showing target detection against sea clutter. The fourth-root law means doubling range requires 16x power.

25 kW
3
100 m²
--
--
--
$$R_{\max} = \left[\frac{P_t G^2 \lambda^2 \sigma}{(4\pi)^3 S_{\min}}\right]^{1/4}$$

Fourth-root law: doubling detection range requires 16x more transmit power.5 Compound K-distributed clutter: sea clutter intensity is modeled as an exponential random variable whose mean is itself Gamma-distributed with shape parameter $\nu$.1 Low $\nu$ (rough sea) produces heavy tails - bright clutter spikes that mimic targets and overwhelm CFAR detectors. High $\nu$ (calm) approaches Rayleigh statistics where detection is straightforward.3,6

Why Maritime Radar is Clutter-Dominated

Maritime radar operates in a fundamentally different regime from free-space or airborne radar. The dominant challenge is signal-to-clutter ratio (SCR), not signal-to-noise.1,3

The sea surface as a reflector. Wind-driven waves, capillary ripples, and swell all backscatter radar energy. At the low grazing angles of ship-borne radar (looking out toward the horizon), even modest seas produce strong returns that fill every range-azimuth cell. A small target's RCS (1-100 m²)7 must compete against clutter returns from an illuminated ocean patch that can span thousands of m². Clutter power typically exceeds thermal noise by 20-40 dB.3

Why not scattering? Atmospheric volume scattering is relatively negligible at typical radar frequencies (S/X-band, 3-10 GHz). Rain and fog cause some attenuation but don't create persistent false targets. The problem is not that the signal is scattered away - it arrives fine - but that it is buried under the surface return from the sea itself.1

Heavy-tailed statistics. Sea clutter follows compound K-distribution statistics (modeled above). The local clutter power varies randomly with heavy tails, producing occasional bright spikes that look exactly like real targets.1,6 This is why simple threshold detectors fail: a fixed threshold that catches targets also triggers on clutter spikes (false alarms), while raising the threshold to avoid false alarms means missing real targets. Switching from Gaussian to K-distribution statistics can degrade detection performance by up to 12 dB.3 CFAR detectors adapt the threshold locally, but still struggle when $\nu$ is low (rough seas) because the tail probabilities are much higher than Gaussian or Rayleigh models predict.

Contrast with other domains. Airborne radar looking down sees targets against a relatively stable ground backdrop.8 Space-based radar tracking satellites operates in a noise-limited regime with no clutter at all.12 Maritime surface radar is uniquely difficult because the clutter source (the ocean) is co-located with the targets of interest, is in constant motion, and exhibits non-Gaussian statistics that defeat classical detection theory.1

SOFAR Channel

Sound speed in water depends on temperature, salinity, and pressure. Temperature drops with depth (less solar heating), while pressure rises (more water above). These opposing gradients create a sound speed minimum at ~1000m - the SOFAR channel - where acoustic energy is trapped by refraction.

Left panel: Three depth profiles - temperature (drops with depth), pressure effect (rises with depth), and their combined result: the sound speed curve with its minimum at the SOFAR axis. Right panel: Animated ray paths from a movable source. Rays near the SOFAR axis refract back toward the minimum and travel far. Rays away from it escape to the surface or seabed and dissipate.

1000 m
100 km
--
--
$$\text{TL}_{\text{SOFAR}} \approx 10\log_{10}(R) + \alpha R \quad \text{vs.} \quad \text{TL}_{\text{normal}} \approx 20\log_{10}(R) + \alpha R$$

Why the minimum exists: Sound speed $\approx 1449 + 4.6T - 0.055T^2 + 0.016D$ (simplified Mackenzie equation,13 where $T$ = temperature in C, $D$ = depth in m). Temperature falls ~1.5C per 100m in the thermocline (0-1000m), pulling speed down. Below the thermocline, temperature stabilizes near 2-4C, and the pressure term ($+0.016D$) dominates, pushing speed back up. The crossing point - typically at 600-1,200m depth - creates the SOFAR minimum.10 Sound refracts toward lower speed, so rays bend back toward the axis - trapping energy in a waveguide. At 1,000 km range, cylindrical spreading provides a ~30 dB advantage over spherical spreading.10

The Information Budget

Every remote sensing system has a fixed rate at which it can extract information from a scene. You must spend that budget across swath width, resolution, and radiometric precision. This tradeoff is not engineering - it is information theory.

A sensor platform has essentially fixed hardware: radiated or received power spectrum over some bandwidth $B$, noise power spectral density $N_0$, platform kinematics (altitude, velocity, look geometry), and total data rate. These determine a maximum Shannon information rate:

$$C = B \log_2\!\left(1 + \frac{S}{N_0 B}\right) \quad \text{bits/second}$$

You then spend that capacity across three dimensions:

W
Swath width (spatial bandwidth)
Δx
Ground resolution (pixel size)
b
Bits per pixel (radiometric precision)

The number of independent cross-track samples per line is $N = W / (2\Delta x)$, where the factor of 2 comes from Nyquist sampling. Each sample carries $b$ bits set by SNR and quantization. If you fix the per-line information budget, then increasing $W$ at fixed $\Delta x$ requires more samples per line - more bits - which forces a reduction elsewhere. The total information-gathering ability is fundamentally bounded.

Information Budget Explorer

Move the sliders to see how swath, resolution, and bits per pixel compete for a fixed information budget. The capacity bar shows utilization - when it overflows, the system cannot physically collect that combination.

100 km
10 m
8 bits
7.5 km/s (LEO satellite)
Information demand -- Mbps
0 Channel capacity: -- Mbps
--
--
--

Why This Hits SAR and Optical Differently

SAR / Radar: Hard Physics Limit
$$W \cdot \delta_{az} \;\geq\; \frac{\lambda R}{2}$$

The pulse repetition frequency (PRF) must be high enough to avoid azimuth ambiguities (Doppler aliasing) but low enough to avoid range ambiguities (echoes from previous pulses returning during the next listen window). These two requirements bound the PRF from both sides, creating an inequality that cannot be violated regardless of engineering investment. It is a property of the waveform itself.

For a LEO SAR at 500 km altitude with X-band ($\lambda$ = 3 cm): $W \times \delta_{az} \geq 7{,}500$ m$^2$. A 1m resolution image can cover at most a 7.5 km swath. A 100 km swath forces resolution to 75m or worse. No amount of power, antenna size, or clever signal processing can break this bound.12,14

EO / Optical: Engineering Challenge
$$\text{SBP} = \left(\frac{D}{\lambda}\right)^2 \times \Omega$$

The space-bandwidth product (SBP) sets the maximum number of independently resolvable spots an optical system can produce: aperture diameter $D$ over wavelength $\lambda$, squared, times the solid-angle field of view $\Omega$. This is the optical equivalent of the SAR swath-resolution inequality.

The difference: you can increase the SBP by building a bigger aperture or using a larger focal plane array. It is an engineering constraint, not a waveform property. A $D$ = 1m telescope at $\lambda$ = 500 nm can resolve ~$4 \times 10^{12}$ spots - enough for 0.5m resolution over a 15 km swath. But you pay in mass, cost, and downlink bandwidth. The information budget still applies - you just have more design freedom in where you hit the wall.12,15

Implication for maritime surveillance

This is why the layered architecture (Part IV of the report) is a consequence of information theory, not an operational preference. OTHR covers 13 million km$^2$ because it uses km-scale resolution - but you cannot tell a tanker from a warship. A SAR satellite achieves 1m resolution but covers only a few thousand km$^2$ per pass. A patrol vessel's EO/IR camera can classify a target's hull markings but covers only the visual horizon. Each tier exists because physics imposes a tradeoff between coverage and capability - and no single platform can be simultaneously wide-area, high-resolution, and persistent.

Sonar Detection

Compare passive and active sonar. Passive listens for a target's own noise (covert, one-way path). Active sends a ping and listens for the echo (detects quiet targets, but two-way path and reveals your position).

120 dB
50 km
75 dB
18 dB
5 dB
--
$$\text{SE} = \text{SL}_{\text{target}} - \text{TL} - (\text{NL} - \text{DI}) - \text{DT}$$

Passive sonar listens for the target's own noise. One-way transmission loss ($R^2$ dependence). Covert - the target doesn't know you're listening. But modern quiet submarines (AIP, electric drive) have reduced source levels by 30-40 dB vs. Cold War boats, shrinking detection ranges from hundreds of km to tens of km.9,10

Active vs Passive: Key Tradeoffs
Passive Active
Path lossOne-way (R2)Two-way (R4) - 16x power to double range
Covert?Yes - target unawareNo - ping reveals your position
Vs. quiet subsStruggles - low SL = short rangeWorks - you supply the energy
OutputBearing only (no range)Range AND bearing (from echo delay)
Limiting factorAmbient noise floorReverberation at short range, noise at long range

Detection Probability vs Range

Probability of detection as a function of range for current sonar settings. Vertical marker shows current range. Note the steeper falloff in active mode from two-way path loss.

Sources

Primary references for the physics, equations, and performance figures used in these visualizations.

  1. Ward, K.D., Tough, R.J.A., Watts, S. Radar Sea Clutter: Scattering, the K Distribution and Radar Performance. IET Digital Library. digital-library.theiet.org
  2. NATO ACT. NATO Mainsail: Multi-Domain Awareness and Insight with AI Layering, 2025. act.nato.int
  3. IET. Maritime Surveillance Radar Part 2: Detection Performance Prediction in Sea Clutter. theiet.org [PDF]
  4. Dark Shipping. SIGINT Dark Shipping. darkshipping.com
  5. Radartutorial.eu. The Radar Range Equation. radartutorial.eu
  6. MathWorks. Maritime Radar Sea Clutter Modeling (MATLAB/Simulink). mathworks.com
  7. Harre, I. RCS in Radar Range Calculations for Maritime Targets. mar-it.de [PDF]
  8. Windward AI. Synthetic Aperture Radar (SAR): The All-Weather Eye on the Oceans. windward.ai
  9. US Navy ES310. Sonar Propagation and Detection Equations. FAS. fas.org
  10. US Navy Undersea Warfare Magazine. SOSUS: The Secret Weapon of Undersea Surveillance. navy.mil
  11. NATO STO. Airborne Maritime Radar Based Submarine Periscope Detection and Discrimination at High Grazing Angles, November 2024. sto.nato.int
  12. IEEE/arXiv. Space-Based Global Maritime Surveillance Part I: Satellite Technologies. arxiv.org
  13. Mackenzie, K.V. Nine-term equation for the sound speed in the oceans. J. Acoust. Soc. Am. 70(3), 807-812, 1981. doi.org
  14. Cumming, I.G., Wong, F.H. Digital Processing of Synthetic Aperture Radar Data. Artech House, 2005. Ch. 3: SAR Signal Properties - PRF constraints and swath-resolution tradeoff.
  15. Goodman, J.W. Introduction to Fourier Optics, 4th ed. W.H. Freeman, 2017. Ch. 6: Space-bandwidth product and information capacity of imaging systems.